Learning stochastic dynamics and predicting emergent behavior using transformers.
Journal
Nature communications
ISSN: 2041-1723
Titre abrégé: Nat Commun
Pays: England
ID NLM: 101528555
Informations de publication
Date de publication:
29 Feb 2024
29 Feb 2024
Historique:
received:
15
03
2022
accepted:
31
01
2024
medline:
1
3
2024
pubmed:
1
3
2024
entrez:
29
2
2024
Statut:
epublish
Résumé
We show that a neural network originally designed for language processing can learn the dynamical rules of a stochastic system by observation of a single dynamical trajectory of the system, and can accurately predict its emergent behavior under conditions not observed during training. We consider a lattice model of active matter undergoing continuous-time Monte Carlo dynamics, simulated at a density at which its steady state comprises small, dispersed clusters. We train a neural network called a transformer on a single trajectory of the model. The transformer, which we show has the capacity to represent dynamical rules that are numerous and nonlocal, learns that the dynamics of this model consists of a small number of processes. Forward-propagated trajectories of the trained transformer, at densities not encountered during training, exhibit motility-induced phase separation and so predict the existence of a nonequilibrium phase transition. Transformers have the flexibility to learn dynamical rules from observation without explicit enumeration of rates or coarse-graining of configuration space, and so the procedure used here can be applied to a wide range of physical systems, including those with large and complex dynamical generators.
Identifiants
pubmed: 38424071
doi: 10.1038/s41467-024-45629-w
pii: 10.1038/s41467-024-45629-w
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
1875Subventions
Organisme : Fonds Wetenschappelijk Onderzoek (Research Foundation Flanders)
ID : V426923N
Organisme : DOE | SC | Basic Energy Sciences (BES)
ID : DE-AC02-05CH11231
Organisme : DOE | Office of Science (SC)
ID : DE-AC02-05CH11231
Organisme : DOE | SC | Basic Energy Sciences (BES)
ID : DE-AC02-05CH11231
Organisme : DOE | Office of Science (SC)
ID : DE-AC02-05CH11231
Informations de copyright
© 2024. The Author(s).
Références
Prinz, J.-H. et al. Markov models of molecular kinetics: Generation and validation. J. Chem. Phys. 134, 174105 (2011).
doi: 10.1063/1.3565032
pubmed: 21548671
Bowman, G. R., Pande, V. S. & Noé, F. An introduction to Markov state models and their application to long timescale molecular simulation, vol. 797 (Springer Science & Business Media, 2013).
Hoffmann, M. et al. Deeptime: a python library for machine learning dynamical models from time series data. Mach. Learn.: Sci. Technol. 3, 015009 (2021).
Wu, H. & Noé, F. Variational approach for learning markov processes from time series data. J. Nonlinear Sci. 30, 23–66 (2020).
doi: 10.1007/s00332-019-09567-y
Mardt, A., Pasquali, L., Wu, H. & Noé, F. Vampnets for deep learning of molecular kinetics. Nat. Commun. 9, 1–11 (2018).
Supekar, R., Song, B., Hastewell, A., Choi, G.P., Mietke, A. & Dunkel, J. Learning hydrodynamic equations for active matter from particle simulations and experiments. Proc. Natl Acad. Sci. 120, e2206994120 (2023).
Maddu, S., Vagne, Q. & Sbalzarini, I. F. Learning deterministic hydrodynamic equations from stochastic active particle dynamics. arXiv preprint arXiv:2201.08623 (2022).
Tsai, S.-T., Kuo, E.-J. & Tiwary, P. Learning molecular dynamics with simple language model built upon long short-term memory neural network. Nat. Commun. 11, 5115 (2020).
doi: 10.1038/s41467-020-18959-8
pubmed: 33037228
pmcid: 7547727
Vaswani, A. et al. Attention is all you need. In Advances in neural information processing systems, 5998–6008 (2017).
Devlin, J., Chang, M.-W., Lee, K. & Toutanova, K. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805 (2018).
Radford, A. et al. Language models are unsupervised multitask learners (2019).
Brown, T. B. et al. Language models are few-shot learners. Advances in neural information processing systems 33, 1877–1901 (2020).
Parmar, N. et al. Image transformer. In International Conference on Machine Learning, 4055–4064 (PMLR, 2018).
Dosovitskiy, A. et al. An image is worth 16x16 words: transformers for image recognition at scale. ICLR (2021).
Levine, S., Kumar, A., Tucker, G. & Fu, J. Offline reinforcement learning: tutorial, review, and perspectives on open problems. arXiv preprint arXiv:2005.01643 (2020).
Wulff, N. & Hertz, J. A. Learning cellular automaton dynamics with neural networks. Adv. Neural Inf. Process. Syst. 5, 631–638 (1992).
Gilpin, W. Cellular automata as convolutional neural networks. Phys. Rev. E 100, 032402 (2019).
doi: 10.1103/PhysRevE.100.032402
pubmed: 31639995
Grattarola, D., Livi, L. & Alippi, C. Learning graph cellular automata. Adv. Neural Inf. Process. Syst. 34, 20983–20994 (2021).
McGibbon, R. T. & Pande, V. S. Efficient maximum likelihood parameterization of continuous-time markov processes. J. Chem. Phys. 143, 034109 (2015).
doi: 10.1063/1.4926516
pubmed: 26203016
pmcid: 4514821
Harunari, P. E., Dutta, A., Polettini, M. & Roldán, É. What to learn from a few visible transitions’ statistics? Phys. Rev. X 12, 041026 (2022).
Frishman, A. & Ronceray, P. Learning force fields from stochastic trajectories. Phys. Rev. X 10, 021009 (2020).
García, L. P., Pérez, J. D., Volpe, G., Arzola, A. V. & Volpe, G. High-performance reconstruction of microscopic force fields from brownian trajectories. Nat. Commun. 9, 1–9 (2018).
Chen, X. Maximum likelihood estimation of potential energy in interacting particle systems from single-trajectory data. Electron. Commun. Probab. 26, 1–13 (2021).
doi: 10.1214/21-ECP416
Campos-Villalobos, G., Boattini, E., Filion, L. & Dijkstra, M. Machine learning many-body potentials for colloidal systems. J. Chem. Phys. 155, 174902 (2021).
Ruiz-Garcia, M. et al. Discovering dynamic laws from observations: the case of self-propelled, interacting colloids. arXiv preprint arXiv:2203.14846 (2022).
Lemos, P., Jeffrey, N., Cranmer, M., Ho, S. & Battaglia, P. Rediscovering orbital mechanics with machine learning. Mach. Learn.: Sci. Technol. 4, 045002 (2023).
Wang, R., Kashinath, K., Mustafa, M., Albert, A. & Yu, R. Towards physics-informed deep learning for turbulent flow prediction. In Proc. 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 1457–1466 (2020).
Wang, R. & Yu, R. Physics-guided deep learning for dynamical systems: A survey. arXiv preprint arXiv:2107.01272 (2021).
Pershin, A., Beaume, C., Li, K. & Tobias, S. M. Training a neural network to predict dynamics it has never seen. Phys. Rev. E 107, 014304 (2023).
doi: 10.1103/PhysRevE.107.014304
pubmed: 36797895
Whitelam, S., Klymko, K. & Mandal, D. Phase separation and large deviations of lattice active matter. J. Chem. Phys. 148, 154902 (2018).
doi: 10.1063/1.5023403
pubmed: 29679965
Gonnella, G., Marenduzzo, D., Suma, A. & Tiribocchi, A. Motility-induced phase separation and coarsening in active matter. Comptes Rendus Physique 16, 316–331 (2015).
doi: 10.1016/j.crhy.2015.05.001
Cates, M. E. & Tailleur, J. Motility-induced phase separation. Annu. Rev. Condens. Matter Phys. 6, 219 (2015).
doi: 10.1146/annurev-conmatphys-031214-014710
O’Byrne, J., Solon, A., Tailleur, J. & Zhao, Y. An introduction to motility-induced phase separation in Out-of-equilibrium Soft Matter, (eds Kurzthaler, C., Gentile, L. & Stone, H. A.) ch. 4, pp. 107–150 (The Royal Society of Chemistry, 2023).
Redner, G. S., Wagner, C. G., Baskaran, A. & Hagan, M. F. Classical nucleation theory description of active colloid assembly. Phys. Rev. Lett. 117, 148002 (2016).
doi: 10.1103/PhysRevLett.117.148002
pubmed: 27740811
Omar, A. K., Klymko, K., GrandPre, T. & Geissler, P. L. Phase diagram of active brownian spheres: crystallization and the metastability of motility-induced phase separation. Phys. Rev. Lett. 126, 188002 (2021).
doi: 10.1103/PhysRevLett.126.188002
pubmed: 34018789
Jang, E., Gu, S. & Poole, B. Categorical reparameterization with gumbel-softmax. arXiv preprint arXiv:1611.01144 (2016).
Zhuang, J. et al. Adabelief optimizer: Adapting stepsizes by the belief in observed gradients. Conf. Neural Inf. Process. Syst. 33, 18795–18806 (2020).
Casert, C., Tamblyn, I. & Whitelam, S. Learning stochastic dynamics and predicting emergent behavior using transformers (2024). https://zenodo.org/doi/10.5281/zenodo.10521014 .